Abstract:
Foundational models are general purpose machine learning models which exhibit broad generalizability and applicability and have led to significant developments in natural language processing and computer vision. These successful examples have spurred the exploration of these foundational models towards the chemical and materials domain in recent years. We will discuss two ongoing directions in the group leveraging foundational models for materials discovery. In one example, we discuss the application of language-based foundational models as intelligent agents for guiding materials discovery. We introduce LLMatDesign, an autonomous language-based framework powered by large language models which can translate human instructions, perform materials design, and evaluate outcomes using provided tools. In the second example, we discuss the application of foundational models towards performing accurate atomistic simulations. We illustrate how these particular models are created via pre-training, and how they can then be fine-tuned to become highly effective interatomic potentials, which are evaluated on a number of simulation benchmarks.
Bio:
Dr. Fung is an Assistant Professor in the School of Computer Science and Engineering at Georgia Tech. Prior to this position, he was a Wigner Fellow at Oak Ridge National Laboratory. A physical chemist by training, Dr. Fung now works at the intersection of scientific artificial intelligence, computing, and materials science/chemistry.
Microsoft Teams Need help?
Join the meeting now
Meeting ID: 215 108 771 271
Passcode: 7s3k66
Dial in by phone
+1 630-556-7958,,439997139# United States, Big Rock
Find a local number
Phone conference ID: 439 997 139#